Trillion LabsOpen weightsReasoning modelAuto-detected

Tri-21B-Think

Name: Tri-21B-Think benchmark results
Creator: BenchLeader

Best configuration ranks #427 of 610 on the BenchLeader Index at 44.0 ±6.7 (thinking reasoning effort).

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Benchmark	thinking	Source
GPQA Diamond (AA)not in index	60.1%#357	Artificial Analysis
Humanity's Last Exam (AA)not in index	5.9%#336	Artificial Analysis

Benchmark	thinking	Source
Terminal-Bench Hard	2.3%#304	Artificial Analysis
τ²-Bench Telecom (AA)not in index	93.3%#44	Artificial Analysis

Benchmark	thinking	Source
AA-Omniscience	-62.3#402	Artificial Analysis

Benchmark	thinking	Source
IFBench	54.6%#148	Artificial Analysis

Benchmark	thinking	Source
AA-LCR	21.7%#353	Artificial Analysis

Benchmark	thinking	Source
AA Intelligence Index	9.6#323	Artificial Analysis